This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug optimization/3329] optimization large memory copies uses kernel memcpy function without user's knowledge.
- From: "dorian dot araneda at intel dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 14 Jul 2003 20:11:36 -0000
- Subject: [Bug optimization/3329] optimization large memory copies uses kernel memcpy function without user's knowledge.
- References: <20010621081601.3329.dorian.araneda@intel.com>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3329
pinskia at physics dot uc dot edu changed:
What |Removed |Added
----------------------------------------------------------------------------
CC|pinskia at physics dot uc |
|dot edu |
dorian dot araneda at intel dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|CLOSED |REOPENED
Resolution|FIXED |
pinskia at physics dot uc dot edu changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|REOPENED |RESOLVED
Resolution| |INVALID
dorian dot araneda at intel dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|RESOLVED |REOPENED
Resolution|INVALID |
pinskia at physics dot uc dot edu changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|REOPENED |RESOLVED
Resolution| |INVALID
dorian dot araneda at intel dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC|dorian dot araneda at intel |
|dot com |
------- Additional Comments From dorian dot araneda at intel dot com 2003-07-14 14:23 -------
is this fixed in gcc 3.2
it is still wrong for gcc to secretly insert a call to memcpy into the
compilation of a closed source operating system indepentent library intended
for use by a kernel mode module. the only way for a kernel module developer to
be aware of this is to either find this bug report or notice the call by chance
in the temp asm file output of the compile.
------- Additional Comments From dorian dot araneda at intel dot com 2003-07-14 14:24 -------
is this fixed in gcc 3.2
it is still wrong for gcc to secretly insert a call to memcpy into the
compilation of a closed source operating system indepentent library intended
for use by a kernel mode module. the only way for a kernel module developer to
be aware of this is to either find this bug report or notice the call by chance
in the temp asm file output of the compile.
------- Additional Comments From pinskia at physics dot uc dot edu 2003-07-14 15:30 -------
Did you not read why this it is okay for gcc to change copy of structs to memcpy, the C
standard says so; so if you do not want a C compiler then do not use gcc, use asm or any
non ANSI C compiler.
------- Additional Comments From dorian dot araneda at intel dot com 2003-07-14 16:19 -------
1. just because ANSI says you can do it does not mean you have to do it.
2. implementation of this ANSI feature is incorrect when it comes to kernel
module development or closed source development. the compler should not
deceptivly insert KERNEL CONFIGURATION UNIQUE dependencies to GPL KERNEL
CONFIGURATION UNIQUE symbols into operating system independent close source
library.
3. ANSI standard does not state that you dont have to add a simple flag to
disable this feature.
the ammount of time to add this comple flag feature would take about a hour-
half hour for the gcc maintainers.
instead, i get anti close source rhetoric that started with Richard Henderson
UNWARRENTED FLAME injected into my converstaion with Alexandre Oliva:
-------------------------------------------------
> This is not intuitive to the normal programmer.
So?
> I have no problem with memcopy being an exported kernel symbol.
> Memcopy may be superduper wonderfuly optimized but I do not want
> it or any other open source symbols in our closed source library.
"memcpy" is hardly an "open source symbol". It's an ISO C
sanctioned library entry point.
I have no sympathy.
---------------------------
------- Additional Comments From dorian dot araneda at intel dot com 2003-07-14 16:24 -------
1. just because ANSI says you can do it does not mean you have to do it.
2. implementation of this ANSI feature is incorrect when it comes to kernel
module development or closed source development. the compler should not
deceptivly insert GPL KERNEL CONFIGURATION UNIQUE symbols dependencies into an
operating system independent close source library.
3. ANSI standard does not state that you dont have to add a simple flag to
disable this feature.
the ammount of time to add this comple flag feature would take about a hour-
half hour for the gcc maintainers.
instead, i get anti close source rhetoric that started with Richard Henderson
UNWARRENTED FLAME injected into my converstaion with Alexandre Oliva:
-------------------------------------------------
> This is not intuitive to the normal programmer.
So?
> I have no problem with memcopy being an exported kernel symbol.
> Memcopy may be superduper wonderfuly optimized but I do not want
> it or any other open source symbols in our closed source library.
"memcpy" is hardly an "open source symbol". It's an ISO C
sanctioned library entry point.
I have no sympathy.
---------------------------
------- Additional Comments From pinskia at physics dot uc dot edu 2003-07-14 17:11 -------
Reread his post, he is saying that since it is required by ISO C standard call, that it cannot
be considered an "open source symbols".
------- Additional Comments From dorian dot araneda at intel dot com 2003-07-14 20:11 -------
whatever.
you dont understand the issue that this "feature" has on linmodems development.
a feature like this should have a manual override to shut it off.
it is people like Richard Henderson who curse linux to an existance of nothing
more than a curiosity thanks to thier 'lack of sympathy' towards linmodem
development.
whats the one feature that MUST be working for a desktop user in the world?
a modem. or else the system is nothing more than a fancy word processor or a
boring single player game box.
and that modem is very likely goign to be a SOFT MODEM. like it or not.
it is not DSL or CABLE or WIRELESS or SATTELITE that connects worldwide linux
users. it is the MODEM and it is going to be that way for a LONG TIME.
so dont come to me with "your code should be OPEN source". i have no control of
that. i dont care that "ANSI says you can do it". there is nothing in that
standard that says you cant put an overide to shut this feature off.
I am just tryign to do my job between a rock and a hard place to
1. improve our modem product
1. improve our linux modem
2. improve linux on a whole for the linux user.
but with people like Henderson it is hard to have sympathy for linux as it
struggles to gain share in the desktop OS segment.
thank you for bringing up this aggravating issue up.
mark it resolved, invalid, close whatever the hell you want to call it.
i dont care anymore.